Comparison of particle size spectrum determination from images made using manual and automated image analysis

A digital processing method is described for determining the size spectra of sub-micron particles in natural water from transmission electron microscopy images of particles collected by ultracentrifugation, and is compared with traditional manual counting and size measurement methods. The processing method is based on the use of the MatLab Image Processing toolbox. The manual method was found to underestimate the population of particles smaller than 40 nm equivalent radius, primarily because of a "fatigue factor" in counting the very large numbers of particles in this size range. By contrast, the manual method produced higher particle counts of particles >50 nm radius, primarily because manual counters tend to group together particles as aggregates that digital processing indicates are not contiguous. The digitally-produced size spectrum for a river water sample was found to closely follow a power-series law down to the smallest particle size.